The (mis)use of big data in Paediatric Anaesthesiology

This dissertation examines the responsible use and potential misuse of big data in paediatric anaesthesiology. Using large multicentre observational databases, it explores how routinely collected clinical data can be used to predict outcomes, generate reference ranges, and answer clinically relevant research questions, while highlighting methodological limitations that may lead to misleading conclusions. The first part evaluates prognostic prediction models for perioperative mortality. A systematic review identified numerous risk scores but found that most lacked robust external validation. Subsequent validation of two leading mortality prediction models using the Multicenter Perioperative Outcomes Group (MPOG) database demonstrated reduced performance compared with original studies and limited clinical utility, particularly given the rarity of mortality in paediatric anaesthesia. The second part focuses on description and association. It presents the first multicentre intraoperative reference values for heart rate and blood pressure in healthy children, investigates associations between intraoperative physiological instability and neurological outcomes in neonates and infants, and evaluates the reliability of administrative databases for studying anaesthesia-related complications. Overall, this dissertation demonstrates that big data can substantially advance paediatric anaesthesia research when used rigorously, but requires careful attention to data quality, bias, validation, clinical relevance, and appropriate statistical methodology to avoid misuse.